Deep-learning model based on retinal images may help assess cardiovascular disease risk
Click Here to Manage Email Alerts
Key takeaways:
- The BioAge model, which uses retinal images, accurately stratified individuals by traditional cardiovascular disease risk biomarkers.
- Leukocyte telomere length shortening was associated with higher BioAge score.
The retinal image-based, deep-learning cardiac BioAge model accurately stratified individuals by traditional cardiovascular disease risk biomarkers, with leukocyte telomere length inversely related to BioAge score, researchers reported.
“It is now recognized that deep learning (DL) can extract data from retinal images to augment the traditional means of estimating atherosclerotic cardiovascular disease risk,” Ehsan Vaghefi, PhD, from the department of optometry and ophthalmology at the University of Auckland in New Zealand, and colleagues wrote in Optometry and Vision Science. “We have developed a DL model (cardiac BioAge) designed to identify those individuals within a tight chronological age band who are at increased risk of cardiac events.”
The BioAge model, based on CLAiR technology, calculates an individual’s 10-year atherosclerotic cardiovascular disease risk score and determines whether it differs from that of their chronological peers.
In this cross-sectional cohort study, Vaghefi and colleagues sought to determine whether results from the model were consistent with other biomarkers of cardiovascular disease risk, as well as leukocyte telomere length.
They identified 33,370 individuals (mean age, 56 years; 55% women) in the UK Biobank with retinal images, leukocyte telomere length measurements and relevant biomarker data, including blood pressure, total cholesterol and HbA1c, who were grouped by sex and ranked by leukocyte telomere length.
According to results, the BioAge model accurately stratified individuals by traditional cardiovascular disease risk biomarkers. Compared with individuals in the bottom quartile of their chronological peer group, those with a cardiac BioAge in the top quartile had significantly higher mean systolic blood pressure, HbA1c and 10-year pooled cohort equation cardiovascular disease risk scores (P < .001).
Further, a higher cardiac BioAge was linked with leukocyte telomere length shortening, a finding that was consistent among both men and women.
“Until now, optometry’s focus has been on ocular disease, but with the emergence of artificial intelligence algorithms, such as our cardiac BioAge DL model, which have been designed specifically to stratify an individual’s risk of systemic disease by examining retinal images, there is now the potential for optometrists to extend the provision of precision medicine beyond their traditional scope of practice,” Vaghefi and colleagues wrote.
“As retinal photographs are routinely captured in optometry practices, these algorithms can be deployed without significant additional investment in primary care, a feature that makes these technologies particularly relevant to low-resource settings.”
Collapse
Click Here to Manage Email Alerts